This post is part of a meta-series. Click here for a list of all posts in this series.
After struggling to get the CarveWright CNC carving machine to cooperate on-and-off over the last year or so, I’m ready to declare defeat on this approach. Every time I think I’ve got “the” problem solved, a new one crops up. First, I had the colorspace issues. Fixed that. Then, I had issues stemming from materials (MDF), so I swapped out for wood. Then I had issues with accuracy, that seemed to stem from the heightmap again. Wasn’t that. These accuracy issues continued to plague. I disassembled the machine, cleaned it, greased it, aligned it, calibrated it, and repeated the whole process numerous times. Once I thought I had it licked, I engaged in an ambitious many-hour carve to get all the pieces finally done…only to discover massive disparity between what I expected and what the final pieces measured, none of which seemed due to the data going in. I sought out advice on the CarveWright user forums, got some new ideas — perhaps I needed to calibrate the machine per board, for each carve in order to achieve the accuracy I sought, for example. But before I could test any of this, new issues appeared — now, boards wouldn’t even measure, complaining that there was a sensor roller error…when he board left the sensor roller because it had fed past it!
That was toward the end of April. The last straw came tonight, when I mustered up the courage to finally see about resolving these issues and test out this per-board calibration hypothesis. I couldn’t get the sensor roller to stop throwing errors, telling the machine to ignore the errors caused different errors to appear, and then — when taking apart the sandpaper belts that feed the board through the machine, I saw that the belts had started to “roll under” themselves again, which was an issue I fixed months ago. It was too much. There are parts I can look into replacing — newer, better; rubber belts instead of sandpaper, for instances — but that costs a great deal of money on top of the money already spent to acquire the machine in the first place (dramatically discounted though it was). I set out to prove that one could make a good-quality stormtrooper helmet on the cheap; this wasn’t that at all and I wasn’t about to keep throwing money at it.
Therefore, I’m changing my approach once again. While the cross-section approach is still something that I think has merit, I’ve come to the point now where I’ve seen enough successful projects that start from naught but paper that I’m going to give that a go. I’ve already got my 3D model, which needs only marginal tweaking to be suitable for that sort of approach, so I should lose little in the accuracy I hoped to achieve with the CarveWright, though I may not end up with a solid wood positive mold that I can pull numerous silicone negatives/poured urethane casts from. Maybe. Who knows, perhaps I will be able to create a mold this way and still use the silicone-and-urethane approach I planned to use all along.
Time to find out.